Fix small bug in semantic seg. multiclass jaccard calculation

BiaPyX · Oct 10, 2024 · 5f9b0aa · 5f9b0aa
1 parent e30485d
commit 5f9b0aa
Show file tree

Hide file tree

Showing 2 changed files with 6 additions and 14 deletions.
diff --git a/biapy/engine/base_workflow.py b/biapy/engine/base_workflow.py
@@ -1739,10 +1739,6 @@ def process_test_sample(self):
                 if self.cfg.MODEL.N_CLASSES > 2 and self.cfg.DATA.TEST.ARGMAX_TO_OUTPUT:
                     _type = np.uint8 if self.cfg.MODEL.N_CLASSES < 255 else np.uint16
                     pred = np.expand_dims(np.argmax(pred, -1), -1).astype(_type)
-                    if self.current_sample["Y"] is not None:
-                        self.current_sample["Y"] = np.expand_dims(np.argmax(self.current_sample["Y"], -1), -1).astype(
-                            _type
-                        )
 
                 # Calculate the metrics
                 if self.current_sample["Y"] is not None:
@@ -1858,10 +1854,6 @@ def process_test_sample(self):
                 if self.cfg.MODEL.N_CLASSES > 2 and self.cfg.DATA.TEST.ARGMAX_TO_OUTPUT:
                     _type = np.uint8 if self.cfg.MODEL.N_CLASSES < 255 else np.uint16
                     pred = np.expand_dims(np.argmax(pred, -1), -1).astype(_type)
-                    if self.current_sample["Y"] is not None:
-                        self.current_sample["Y"] = np.expand_dims(np.argmax(self.current_sample["Y"], -1), -1).astype(
-                            _type
-                        )
 
                 if self.cfg.TEST.POST_PROCESSING.APPLY_MASK:
                     pred = apply_binary_mask(pred, self.cfg.DATA.TEST.BINARY_MASKS)

diff --git a/biapy/engine/metrics.py b/biapy/engine/metrics.py
@@ -183,12 +183,12 @@ def __call__(self, y_pred, y_true):
             y_true = torch.cat((1 - y_true, y_true), 1)
 
         if self.num_classes > 2:
-            return self.jaccard(
-                y_pred,
-                (y_true.squeeze() if y_true.shape[0] > 1 else y_true.squeeze().unsqueeze(0)),
-            )
-        else:
-            return self.jaccard(y_pred, y_true)
+            if y_pred.shape[1] > 1:
+                y_true = y_true.squeeze()
+            if len(y_pred.shape)-2 == len(y_true.shape):
+                y_true = y_true.unsqueeze(0)
+
+        return self.jaccard(y_pred, y_true)
 
 
 class instance_metrics: